search

junyongyou / triq

TRIQ implementation
MIT License
134 stars 23 forks source link

About dataset #9

Closed zmm96 closed 3 years ago

zmm96 commented 3 years ago

Hello, I have a question, the data shape of koniq-10k dataset is not consistent. Some is (224,224), otherwise some is(224,224,3)。but I do not find the process about the difference. Can you tell me more about the detail? thanks a lot.

junyongyou commented 3 years ago

Hello, I have a question, the data shape of koniq-10k dataset is not consistent. Some is (224,224), otherwise some is(224,224,3)。but I do not find the process about the difference. Can you tell me more about the detail? thanks a lot.

If I remember correctly, I should have just simply replicated a 224224 image to 224224*3 image by repeating the single channel. I should have done this before training the model, i.e., writing the 3-channel image into a new file. So such process was not included in the code.

zmm96 commented 3 years ago

I have one more question. the code　you use to process the data before training is mics/imagenet_handler or databases/random_split_imagesset？

junyongyou commented 3 years ago

I have one more question. the code　you use to process the data before training is mics/imagenet_handler or databases/random_split_imagesset？

I think these two should be used jointly. However, I would recommend you to do train-val-test split yourself, just use the TRIQ model implemented here. If you only want to test the model, you can use my trained weights.

zmm96 commented 3 years ago

can you tell me more about the detail　in　mics / imagenet_handler　？

junyongyou commented 3 years ago

can you tell me more about the detail　in　mics / imagenet_handler　？

Hi, I think I have put sufficient comments in each method. Please be specific if you have any questions.